Occasional DNS failures on fleeting runners
We've see occasional failures in resolving the internal DNS clone host: us-east1-c.ci-gateway.int.gprd.gitlab.net
Getting source from Git repository 00:06
Fetching changes...
Initialized empty Git repository in /builds/gitlab-org/gitlab-runner/.git/
Created fresh repository.
fatal: unable to access 'https://us-east1-c.ci-gateway.int.gprd.gitlab.net:8989/gitlab-org/gitlab-runner.git/': Could not resolve host: us-east1-c.ci-gateway.int.gprd.gitlab.net
Uploading artifacts for failed job
We've seen this on fleeting runners:
- https://gitlab.com/gitlab-org/gitlab-runner/-/jobs/6693416776
- https://gitlab.com/gitlab-org/gitlab-runner/-/jobs/6815706052
I wonder if there is some race condition with cloud-init
as with #205? For example, I wonder if the /etc/systemd/system/docker.service.d/05-iptables-restore-wants.conf
doesn't always get loaded, though I'm not sure if that would affect this.
/cc: @ajwalker @tmaczukin
Edited by Stan Hu